[Mageia-dev] Finalizing update process

[Colin Guthrie]

'Twas brillig, and Ahmad Samir at 08/06/11 22:48 did gyre and gimble:
> On 8 June 2011 23:38, Stew Benedict <stewbintn at gmail.com> wrote:
>> On 06/08/2011 05:31 PM, Michael Scherer wrote:
>>>
>>> Le mercredi 08 juin 2011 à 19:48 +0300, Ahmad Samir a écrit :
>>>>
>>>> On 8 June 2011 18:44, Michael Scherer<misc at zarb.org>  wrote:
>>>>>
>>>>> Le mercredi 08 juin 2011 à 10:40 +0200, Anne nicolas a écrit :
>>>>>>
>>>>>> Hi there
>>>>>>
>>>>>> We have some stuff to complete here:
>>>>>> http://mageia.org/wiki/doku.php?id=security
>>>>>>
>>>>>> <http://mageia.org/wiki/doku.php?id=security>Can we spend the 2 or 3
>>>>>> coming
>>>>>> days to finalize it and start updates submits?
>>>>>
>>>>> Pascal is working on this.
>>>>>
>>>>> So here is a proposal :
>>>>> - anybody can submit a package to updates_testing.
>>>>> - once submitted to testing, it should ask to QA to test, along with :
>>>>>  - a reason for the update ( likely bug number )
>>>>>  - potentially a priority ( ie, if this is just a translation update or
>>>>> a urgent 0 day exploit )
>>>>>  - a way to test the bug and see it is fixed
>>>>>  - text for the update
>>>>>
>>>>> - qa validate the update ( with process to define )
>>>>>
>>>>> - someone move the package from updates_testing to testing
>>>>
>>>> Isn't it cleaner to rebuild when submitting to */updates? instead of
>>>> moving.
>>>
>>> Well, that depend on the way package is built. In fact, it should not
>>> matter much, as we should not do a change that could break a existing
>>> software in update, and so this should be the same in updates_testing
>>> ( ie, they should be the same wrto ABI )
>>>
>>> But that's also why I ask here :)
>>>
>>>
>> If you're going to rebuild *after* QA, you've just invalidated your QA.
>> (yeah, I know it *should* be the same, but stuff happens)
>>
> 
> You're right (even if that's never happened for 3-4 years in mdv,
> since sec team rebuilt the packages when pushing to */updates IIRC).


Personally, and this might just be me, I always submit my packages to
*testing with a subrel of 0.1, 0.2 0.3 etc etc. Users then test my
various iterations. When I'm happy and when it's ready to pass to QA, I
set the subrel to 1. This way the final version that should hit updates
is nice and neat.

In an ideal world, QA would validate it for me then change the subrel
for me. That process would require a rebuild.

I'm not sure what others feel about this? It's not impossible to just do
this as a matter of course as part of the process we go through and
increment subrel to a round number before handing over to QA... although
maybe I'm just a bit too anal about neat version numbers :p

Col




-- 

Colin Guthrie
mageia(at)colin.guthr.ie
http://colin.guthr.ie/

Day Job:
  Tribalogic Limited [http://www.tribalogic.net/]
Open Source:
  Mageia Contributor [http://www.mageia.org/]
  PulseAudio Hacker [http://www.pulseaudio.org/]
  Trac Hacker [http://trac.edgewall.org/]